OpenClaw Slack Integration: Deploy Your AI Agent to Slack

Why Slack for AI Agents

Slack has over 750,000 organizations using it daily. Unlike standalone dashboards or CLI tools, a Slack-based agent meets your team where they already work. There is no context switch. Someone asks a question in a channel, the agent responds in seconds, and the entire team sees the answer.

Practical use cases for OpenClaw agents in Slack include: answering recurring questions from a knowledge base, generating daily reports and posting them to a channel, monitoring external services and alerting when something breaks, triaging incoming support requests by analyzing the message and routing to the right person, and running SQL queries against a database when someone types a slash command.

The Slack integration uses Slack's Socket Mode by default, which means you do not need a public URL or webhook endpoint. Your OpenClaw gateway connects outbound to Slack's servers over a WebSocket, making it work behind firewalls and NATs without any additional networking configuration.

Prerequisites

Before you start, make sure you have the following ready:

Step 1: Create a Slack App

Every Slack bot starts as a Slack App. This is the container that holds your bot user, permissions, event subscriptions, and tokens. The setup takes about three minutes.

Step 2: Configure Bot Permissions (OAuth Scopes)

Slack uses OAuth scopes to control what your bot can and cannot do. Go to "OAuth & Permissions" in the sidebar and scroll down to "Bot Token Scopes." Add the following scopes:

After adding scopes, scroll up and click "Install to Workspace" (or "Reinstall to Workspace" if you already installed it before adding scopes). Authorize the app. You will see a "Bot User OAuth Token" starting with xoxb-. Copy this token.

Subscribe to Events

Go to "Event Subscriptions" in the sidebar and toggle it on. Under "Subscribe to bot events," add these events:

# Required bot events for OpenClaw Slack integration

app_mention        # Triggers when someone @mentions your bot
message.channels   # Messages in public channels where bot is present
message.im         # Direct messages to the bot

# Optional events
message.groups     # Messages in private channels
message.mpim       # Messages in multi-person DMs

Step 3: Configure OpenClaw for Slack

With your Slack app created and tokens in hand, configure OpenClaw to connect to it. You need two tokens: the Bot User OAuth Token (xoxb-) and the App-Level Token (xapp-) for Socket Mode.

# Required: Bot User OAuth Token (starts with xoxb-)
openclaw config set channels.slack.bottoken xoxb-YOUR-BOT-TOKEN

# Required: App-Level Token for Socket Mode (starts with xapp-)
openclaw config set channels.slack.apptoken xapp-YOUR-APP-TOKEN

# Optional: Reply in threads instead of the main channel
openclaw config set channels.slack.replyInThread true

# Optional: Only respond when @mentioned (ignore other messages)
openclaw config set channels.slack.mentionOnly true

# Optional: Restrict to specific channel IDs
openclaw config set channels.slack.allowedChannels '["C01ABC123", "C02DEF456"]'

# Optional: Set a custom status emoji
openclaw config set channels.slack.statusEmoji ":robot_face:"

# Start or restart the gateway
openclaw gateway restart

After restarting the gateway, check the logs to confirm the connection:

# Verify Slack connection
openclaw gateway logs | grep slack

# Expected output:
# [channel:slack] connected to workspace "YourWorkspace"
# [channel:slack] socket mode active, listening for events
# [channel:slack] bot user: @your-bot-name (U01XXXXXXXX)

If you see [channel:slack] connected in the logs, your agent is live. Go to any channel where the bot is present, type @YourBot hello, and you should get a response.

Step 4: Write a SOUL.md Optimized for Slack

Your SOUL.md defines how the agent behaves. When targeting Slack, there are specific considerations: messages should be concise (Slack is not the place for essays), formatting should use Slack's mrkdwn syntax (not standard Markdown), and the agent should understand channel context.

# Support Agent — SOUL.md

## Identity
You are a technical support agent for the engineering team.
Your name is Atlas. You respond in Slack channels and DMs.

## Communication Rules
- Keep responses under 300 words. Slack messages should be scannable.
- Use Slack mrkdwn formatting: *bold*, _italic_, `code`, ```code blocks```.
- Use bullet points for lists, not numbered lists (they render poorly in Slack).
- When sharing links, use Slack format: <https://example.com|Display Text>.
- Never use Markdown headers (# H1) — they do not render in Slack.
- For long responses, suggest continuing in a thread.

## Behavior
- If asked about something outside your knowledge, say so. Do not guess.
- When a question is ambiguous, ask one clarifying question before answering.
- Tag relevant team members with <@USER_ID> when escalation is needed.
- React with :eyes: when you start processing a complex request.

## Skills
- Search the internal knowledge base at https://docs.internal.company
- Query the PostgreSQL database for user and order data
- Check service health status via the monitoring API

## Boundaries
- Never share API keys, tokens, or credentials in channel messages.
- Never modify production data. Read-only access to databases.
- Escalate billing and account deletion requests to <@U01ADMIN123>.

The key differences from a generic SOUL.md: explicit Slack formatting rules, message length limits appropriate for chat, instructions for using Slack-specific features like user mentions and reactions, and clear escalation paths using Slack user IDs.

Environment Variables

If your agent uses skills that require API keys or database connections, configure them as environment variables in your OpenClaw project:

# Slack tokens (set via openclaw config, but can also be env vars)
SLACK_BOT_TOKEN=xoxb-your-bot-token
SLACK_APP_TOKEN=xapp-your-app-token

# Agent-specific environment variables
DATABASE_URL=postgresql://user:pass@host:5432/mydb
KNOWLEDGE_BASE_API_KEY=kb_xxxxxxxxxxxx
MONITORING_API_URL=https://status.internal.company/api

# OpenAI / Anthropic model key (used by the agent)
OPENAI_API_KEY=sk-xxxxxxxxxxxxxxxx

Step 5: Test Your Slack Agent

Before rolling the agent out to your whole team, test it thoroughly. Here is a testing checklist:

# Watch logs in real-time while testing
openclaw gateway logs --follow | grep slack

# Check for errors specifically
openclaw gateway logs --lines 50 | grep -i "error|failed|timeout"

# Verify message flow
# Expected log pattern:
# [channel:slack] received message from U01USER123 in C01CHANNEL
# [agent] processing message: "what can you do?"
# [agent] response generated (247 tokens, 1.3s)
# [channel:slack] sent response to C01CHANNEL

Advanced: Slash Commands

Slash commands give your agent a structured interface. Instead of free-form messages, users type /ask how do I reset my password? and get a clean response. This is especially useful for agents that handle specific tasks like database queries, report generation, or ticket creation.

Register the Command

In your Slack app settings, go to "Slash Commands" and click "Create New Command."

# Command: /ask
# Request URL: https://your-host:18789/slack/commands
#   (or use Socket Mode — no URL needed)
# Short Description: Ask the AI agent a question
# Usage Hint: [your question here]

# Command: /report
# Request URL: https://your-host:18789/slack/commands
# Short Description: Generate a report
# Usage Hint: [daily|weekly|monthly] [metric-name]

# Command: /query
# Request URL: https://your-host:18789/slack/commands
# Short Description: Run a database query
# Usage Hint: [SQL query or natural language question]

Handle Commands in OpenClaw

Configure OpenClaw to route slash commands to your agent:

# Enable slash command handling
openclaw config set channels.slack.enableCommands true

# Map commands to specific agent behaviors
openclaw config set channels.slack.commands '{
  "/ask": {
    "agent": "support-bot",
    "ephemeral": false
  },
  "/report": {
    "agent": "metrics-bot",
    "ephemeral": true
  },
  "/query": {
    "agent": "data-bot",
    "ephemeral": true
  }
}'

# Restart to apply
openclaw gateway restart

The ephemeral flag controls visibility. When set to true, only the user who ran the command sees the response. This is ideal for sensitive data like database queries or personal reports. When false, the response is visible to everyone in the channel.

Troubleshooting

Here are the most common issues you will encounter and how to fix them.

# Check if the gateway is receiving events
openclaw gateway logs --follow | grep "slack.*received"

# If no events appear, the issue is on the Slack side:
# 1. Verify event subscriptions are saved
# 2. Reinstall the app to the workspace
# 3. Make sure Socket Mode is enabled
# 4. Check the App-Level Token is correct

# Verify your tokens
openclaw config get channels.slack.bottoken
# Should start with xoxb-

openclaw config get channels.slack.apptoken
# Should start with xapp-

# If wrong, update them:
openclaw config set channels.slack.bottoken xoxb-CORRECT-TOKEN
openclaw config set channels.slack.apptoken xapp-CORRECT-TOKEN
openclaw gateway restart

# Invite the bot to a channel (run in Slack)
/invite @YourBot

# If using allowedChannels, make sure the channel ID is listed
openclaw config get channels.slack.allowedChannels

# Get a channel's ID: right-click channel name > Copy > Copy Link
# The ID is the last segment: https://workspace.slack.com/archives/C01ABC123

# Add a response delay to reduce API call frequency
openclaw config set channels.slack.responseDelay 1500

# Limit concurrent message processing
openclaw config set channels.slack.maxConcurrent 2

# Check rate limit hits in the logs
openclaw gateway logs | grep "rate.limit|429"

# Enable reconnection with backoff
openclaw config set channels.slack.autoReconnect true
openclaw config set channels.slack.reconnectDelay 5000
openclaw config set channels.slack.maxReconnectAttempts 15

# Use pm2 for production uptime
pm2 start "openclaw gateway start" --name openclaw-slack
pm2 save && pm2 startup

Skip the Setup with CrewClaw

If you want a Slack-connected AI agent without managing tokens, Socket Mode, event subscriptions, and gateway processes yourself, CrewClaw gives you a pre-configured agent with Slack integration included. Pick a role, customize the SOUL.md in the visual builder, and get a complete deployment package with Slack channel configuration, Docker setup, and environment templates. The entire process takes about 60 seconds.

Related Guides

Frequently Asked Questions